Measurement Without a Decision Is a Hobby

I built a tracking spreadsheet and a protocol page with three columns of target ranges. So take this as a counterweight to my own work rather than a swipe at anyone else's.

Most tracking is a hobby wearing the costume of rigour.

#The statistical problem

Individual biomarker fluctuations, measured frequently in one person, will generate apparent trends that are indistinguishable from noise.

That isn't a claim about discipline. It's arithmetic. Every measurement has biological variation and analytical variation. Sample often enough and you will see runs — three readings drifting up, a dip after you changed something — that are entirely consistent with a flat underlying value.

And here's the part that should be uncomfortable: dense measurement makes this worse, not better. More samples means more opportunities for a random run to look like a signal. Without a pre-specified analysis, higher frequency buys you more false patterns, not more truth.

This is the specific criticism credible aging researchers level at high-profile self-experimenters — that n=1 with a hundred simultaneous variables and daily readings cannot separate signal from noise, however good the instruments are.

#The attention problem

A number you look at daily and never act on has a cost, and it isn't zero.

HRV is the clearest case. It's a real physiological measure. But most people reading a daily readiness score have no pre-committed rule for what a low one changes — they train anyway, or they skip and would have skipped regardless. The number didn't route into a decision; it just added a small negative emotion to the morning.

Meanwhile the thing the number was proxying for — did you sleep, do you feel wrecked — was already available for free.

#The rule that fixes it

Before adding any metric, answer one question: what value would change what I do?

If you can't state the threshold and the action in advance, the metric isn't measurement. It's monitoring, and monitoring without a response is just watching.

Applied honestly, this kills most of what people track:

#The version I'll defend

Measurement earns its place when it is infrequent, decision-linked, and hard to fake.

One blood panel a year beats a year of daily readiness scores, because it produces a number you act on annually and the other produces 365 numbers you act on never. A logged training set beats a wearable estimate, because it's the actual thing rather than a model of it.

And the tracker I built is defensible for exactly one reason: every column on it feeds a decision. Weight feeds the intake adjustment. Sets feed progressive overload. The panel feeds the annual review. If I'd added an HRV column it would have been decoration.

#The uncomfortable summary

The protocol pages on this site are full of numbers, and I stand behind them. But the ranking of what actually moves outcomes is roughly:

1. Doing the thing
2. Doing the thing consistently
3. Measuring whether the thing worked, occasionally
4. Measuring continuously

Most people invert three and four, then wonder why the graphs are so detailed and the results aren't.

---

This comes out of an ongoing protocol. The research behind it — every mechanism, every source tier, and an append-only record of everything it has gotten wrong — is at /protocol.